EP1650979A1 - Method and system for demosaicing artifact removal - Google Patents

Abstract

Color image signals, as derived e.g. by interpolating the output from a color filter array (CFA) are arranged in pixels, each pixel having associated detected (10) color information for a first color as well as undetected filled-in (20) color information for at least a second and a third color. The images are thus exposed to false color and zipper effect artifacts, and are subject to processing preferably including the steps of:

• checking (30) the images for the presence of zipper effect artifacts, and
  • i) if said checking (30) reveals the presence of zipper effect artifacts, applying a zipper effect removal process (50) to said image signals;
  • ii) if said checking (30) fails to reveal the presence of zipper effect artifacts, applying a false color removal process (40) to said image signals.

False color and zipper effect artifacts are thus preferably both reduced by adaptively using the zipper effect removal process (50) and the false color removal process (40).

Description

Field of the invention
• The invention relates to image/video processing techniques.
• Specifically, the invention relates to de-mosaicing artifact removal, and was developed by paying attention to the.possible application to digital cameras. Reference to this possible field of application is not however to be construed in a limiting sense of the scope of the invention.
Description of the related art
• Digital still cameras include two-dimensional image sensing arrays, such as a Charge Coupled Device (CCD) that integrates incident scene light over a predetermined time to provide an electronic information signal corresponding to the scene light intensity incident to the array. Such a two-dimensional image-sensing array comprises a predetermined number of discrete image sensing elements or pixels, which respond to incident illumination to provide an electronic information signal corresponding to the intensity of the incident illumination.
• In order to record color images through a single sensor, the surface of the sensor is covered with a mosaic of colored filters. This kind of sensor is called a Color Filter Array (CFA). In essence, each pixel has its own spectrally selective filter to peer through. The most popular color filter array pattern used today is the Bayer Pattern, which is shown in Figure 1.
• Since each image sensing element can only detect one color of illumination (R=Red, G=Green, B=Blue), the color information for the other colors not detected by that image sensing element must be filled in; thus achieving a "de-mosaicing" effect, which is usually achieved by interpolation.
• Conventional types of interpolation, however, can provide images with objectionable aliasing artifacts such as "color fringes" near sharp edges. There are two main types of de-mosaicing artifacts, named false colors and zipper effect. False colors are those artifacts corresponding to noticeable color errors as compared to the original, non-mosaiced image. The zipper effect refers to abrupt or unnatural changes of color differences between neighboring pixels, manifesting as an "on-off" pattern.
• A conventional approach to solve these problems is to eliminate the color fringes at the expenses of image sharpness by blurring the picture, so that the edges are not sharp enough to create a color fringe. Blurring the image in this manner, however, has its obvious disadvantages resulting in a reduction in resolution. Therefore, it is necessary to provide a de-mosaicing artifact removal technique that reduces color fringing without the amount of blurring otherwise required.
• A technique to resolve color fringes without blurring the images is proposed in US-A-4 724 395.
• The approach disclosed in US-A-4 724 395 stems from the consideration that natural images exhibit a strong correlation of the red, green and blue channels, especially for high frequencies, so that they are likely to have the same texture and edge locations.
• Because of this inter-channel correlation, the difference between two colors in a neighborhood is nearly constant, while it rapidly increases and decreases in the area of sharp gray edges, where color interpolation has introduced false colors. The approach of US-A-4 724 395 applies a median filter to the inter-channel differences in a neighborhood and then uses the resulting value to force pixels with distinct colors to be more similar to their neighbors, avoiding false colors. Moreover, the Color Filter Array color value at each pixel is not altered.
• In general, the approach of US-A-4 724 395 is rather effective in suppressing de-mosaicing artifacts, while preserving sharp edges. However, some de-mosaicing artifacts still remain around sharp edges and fine details. This is partly due to the fact that each pixel has independent inter-channel differences, and filtering the differences separately does not take into account the spectral correlation between color planes.
• The arrangement disclosed in Wenmiao Lu, Yap-Peng Tan, "Color Filter Array De-mosaicing: New Method and Performance Measures", IEEE Transactions on Image Processing, vol. 12, no. 10, October 2003, incorporates median filtering with the spectral correlation for more effective suppression of de-mosaicing artifacts by lifting the constraint of keeping the original CFA-sampled color values intact. Furthermore, Lu et al. make use of the latest processed color values to filter the subsequent pixels so that estimation errors can be effectively diffused into local neighborhoods.
• To sum up, the approach of US-A-4 724 395 does not remove all the artifacts introduced by conventional color interpolation techniques, and moreover, under certain conditions, produces an annoying zipper effect in the horizontal and vertical directions.
• Conversely, the approach of Lu et al. removes more extensively false colors and artifacts in comparison with the approach of US-A-4 724 395, but considerably blurs the images, because it adjusts the green channel of each pixel by means of an average of both the red and blue values of the same pixel.
Object and summary of the invention
• In view of the foregoing, the need is felt for an improved arrangement that may dispense with the disadvantages of the prior art arrangements discussed in the foregoing.
• The object of the invention is thus to provide a fully satisfactory response to that need.
• According to the present invention, that object is achieved by means of a method having the features set forth in the claims that follow, these claims being an integral part of the disclosure of the present invention. The invention also relates to a corresponding system as well as a related computer program product, loadable in the memory of at least one computer and including software code portions for performing the steps of the method of the invention when the product is run on a computer. As used herein, reference to such a computer program product is intended to be equivalent to reference to a computer-readable medium containing instructions for controlling a computer system to coordinate the performance of the method of the invention. Reference to "at least one computer" is evidently intended to highlight the possibility for the present invention to be implemented in a distributed/modular fashion.
• A preferred embodiment of the invention provides for processing image signals arranged in pixels, each pixel having associated detected color information for a first color as well as undetected, filled-in (e.g. interpolated) color information for at least a second and a third color, whereby said images are exposed to false color and zipper effect artifacts. Such preferred embodiment of the invention includes the steps of:
• checking said images for the presence of zipper effect artifacts, and
  • - i) if said checking reveals the presence of zipper effect artifacts, applying a zipper effect removal process to said image signals;
  • - ii) if said checking fails to reveal the presence of zipper effect artifacts, applying a false color removal process to said image signals, whereby said false color and zipper effect artifacts are both reduced by adaptively using said zipper effect removal process and said false color removal process.
• Such an adaptive approach, whereby false color and zipper effect artifacts are both reduced by adaptively using said zipper effect removal process and said false color removal process is held to constitute a presently preferred embodiment of the invention. However the invention is in no way limited to such preferred embodiment, and also encompasses arrangements wherein only one of said zipper effect removal process and said false color removal process is performed.
Brief description of the annexed drawings
• The invention will now be described, by way of example only, with reference to the enclosed figures of drawing, wherein:
• Figure 1 has been already described in the foregoing;
• Figure 2 is a schematic block diagram of a de-mosaicing artifact removal technique described herein; and
• Figure 3 is a detailed block scheme of the arrangement described herein.
Detailed description of preferred embodiments of the invention
• The main operating principles of the exemplary arrangement described herein (which corresponds to the presently preferred embodiment specifically mentioned in the foregoing) can be expressed as follows:
• the technique is assumed to be applied on a set of pixels;
• for each pixel, the color channel whose sample is available from the CFA (e.g. Bayer) pattern is left unchanged;
• inter channel correlation is taken into account, considering two inter-channel differences (red-green and blue-green) median filtered over a given neighborhood and using these two differences together to filter pixels;
• these latest processed color values are used to filter the subsequent pixels;
• false colors and zipper effect are both reduced by adaptively using two different blocks.
• A block diagram representing the de-mosaicing artifact removal technique described herein is shown in Figure 2.
• In Figure 2, reference number 10 designates an image sensor adapted to generate an image signal arranged in e.g. a Bayer pattern (or, notionally, any other CFA pattern); in the same Figure, reference 20 designates an interpolator block, and reference 30 designates a zipper detector block.
• The zipper detector block 30 operates according to a given zipper effect characterization to produce a control signal which enables a false color removal block 40 and/or a zipper effect removal block 50. For ease of description the blocks 40 and 50 are shown herein as totally independent modules. It will however become apparent from the following description that these modules may advantageously share certain sub-modules, i.e. being jointly configured for performing operations that are common to the false color removal process and the zipper effect removal process described in detail in the following.
• The zipper detector block 30 produces its output control signal by verifying if those conditions that determine zipper effect are satisfied or not.
• The zipper effect is due to sudden variations of color differences between neighboring pixels.
• As an example, assume that the central pixel of a local window belongs to a dark horizontal line within a light homogeneous region. Both the blue-green and the red-green differences along the line will be very different from those calculated for the rest of the mask. Therefore, if the missing color channels at the central pixel are reconstructed through these wrong differences, an "on-off" pattern will be generated.
• More specifically, zipper effect arises when the following three conditions are satisfied:
• along the horizontal (vertical) direction passing through the central pixel, the inter-channel difference between the green channel and the other one (for which we have the original sampled values) is almost constant;
• along the vertical (horizontal) direction the inter-channel difference between the green channel and the other one (for which we have the original sampled values) is not constant;
• the trend of the central pixel channel in the vertical (horizontal) direction is increasing or decreasing, or there is a minimum or a maximum in the central pixel.
• Preferably, these three conditions are evaluated on a 5x5 mask, but the dimensions of such a mask are not per se critical: for instance it is also possible to use a 3x3 mask.
• If these conditions are satisfied, the zipper detector block 30 enables the zipper effect removal algorithm (block 50); otherwise it enables the false colors removal algorithm (block 40).
• The first two conditions state that inter-channel differences are almost constant along a direction and not constant in the orthogonal one. In order to verify both these two conditions, the similarity among inter-channel differences has been evaluated according to a given threshold. An increase in the value of this threshold causes the growth of the amount of removed zipper effect. However, the three causes of the zipper effect could be satisfied by pixels belonging to detailed regions of an image, leading to a reduction in details. A trade off between zipper effect reduction and detail preservation should be considered by appropriately setting the threshold.
• The third condition states that the trend of the central pixel channel, along the direction having not constant inter-channel differences, follows a precise law. In particular, this trend:
• is increasing or decreasing if the pixel belongs to a line separating a dark region and a light one;
• has a minimum if the pixel belongs to a dark line within a light background;
• has a maximum if the pixel belongs to a light line within a dark background.
• The false colors removal block 40 adjusts the three color values at the central pixel of a local window distinguishing the following two cases, according to the original sampled channel of the central pixel:
• G case $$\begin{array}{l}{\widehat{\mathrm{H}}}_{\mathrm{CENTER}}={\mathrm{G}}_{\mathrm{CENTER}}+{\mathrm{v}}_{\mathrm{H}\mathrm{G}}\\ {\widehat{\mathrm{G}}}_{\mathrm{CENTER}}={\mathrm{G}}_{\mathrm{CENTER}}\\ {\widehat{\mathrm{J}}}_{\mathrm{CENTER}}={\mathrm{G}}_{\mathrm{CENTER}}+{\mathrm{v}}_{\mathrm{J}\mathrm{G}}\end{array}$$
  
• R/B case (e.g. H) $$\begin{array}{l}{\widehat{\mathrm{H}}}_{\mathrm{CENTER}}={\mathrm{H}}_{\mathrm{CENTER}}\\ {\widehat{\mathrm{G}}}_{\mathrm{CENTER}}={\mathrm{H}}_{\mathrm{CENTER}}-{\mathrm{v}}_{\mathrm{H}\mathrm{G}}\\ {\widehat{\mathrm{J}}}_{\mathrm{CENTER}}={\mathrm{H}}_{\mathrm{CENTER}}-{\mathrm{v}}_{\mathrm{H}\mathrm{G}}+{\mathrm{v}}_{\mathrm{J}\mathrm{G}}\end{array}$$
  
• where: $${\mathrm{v}}_{\mathrm{H}\mathrm{G}}=\mathrm{median}\left\{{\mathrm{H}}_{\mathrm{i}\mathrm{j}}-{\mathrm{G}}_{\mathrm{i}\mathrm{j}}|\left(\mathrm{i},\mathrm{j}\right)\in \mathrm{\Re }\right\},$$

  

  $${\mathrm{v}}_{\mathrm{J}\mathrm{G}}=\mathrm{median}\left\{{\mathrm{J}}_{\mathrm{i}\mathrm{j}}-{\mathrm{G}}_{\mathrm{i}\mathrm{j}}|\left(\mathrm{i},\mathrm{j}\right)\in \mathrm{\Re }\right\}$$

  
• G is green (G), H is red (R) or blue (B) and J is blue (B) or red (R), respectively, and

  

  is the support of the median filter that covers 3x3 pixels. Alternatively, a local window with different dimensions, such as e.g. 5x5 pixels, can be used.
• As in the technique proposed by Lu et al. discussed in the introductory portion of this description, the color values of the central pixel are replaced by Ĥ_CENTER, Ĝ_CENTER and Ĵ_CENTER, so that they will be involved in filtering the following pixels, spreading the estimation errors into local neighborhoods.
• Additionally, it will be appreciated that, in direct comparison with the arrangement of US-A-4 724 395, in the arrangement described herein for those pixels which receive only red or blue light, i.e. the blue and red color, respectively, the information concerning the "missing" colors (and specifically the information concerning the blue color for those pixels which receive only red light, and the information concerning the red color for those pixels which receive only blue light) is computed by using both median filtered inter-channel differences (v_HG and v_JG), and hence the spectral correlation between color planes is taken into account. Thanks to this solution this technique efficiently eliminates false colors without blurring the image.
• The false colors removal block 40 computes both the two missing colors by using the median filtered inter-channel differences. The zipper effect removal block 50, instead, calculates one missing color as an average of some surrounding pixels of the central pixel. In detail, these pixels (preferably derived from the CFA - e.g. Bayer - pattern produced by the sensor 10) are chosen along the direction that has almost constant inter-channel differences. The other missing color is computed by adding, to the reconstructed green channel, the appropriate median filtered inter-channel difference.
• We have to distinguish two different cases, according to the original sampled color of the central pixel (G case and R/B case).
• The two equations for zipper effect suppression can be written as follows:
• G case $$\begin{array}{l}{\hat{\mathrm{G}}}_{\mathrm{CENTER}}={\mathrm{G}}_{\mathrm{CENTER}}\\ {\hat{\mathrm{H}}}_{\mathrm{CENTER}}=\mathrm{f}\left({\mathrm{H}}_{\mathrm{SURROUND}}\right)\\ {\hat{\mathrm{J}}}_{\mathrm{CENTER}}={\mathrm{G}}_{\mathrm{CENTER}}+{\mathrm{v}}_{\mathrm{J}\mathrm{G}}\end{array}$$
  
• R/B case (e.g. H) $$\begin{array}{l}{\hat{\mathrm{G}}}_{\mathrm{CENTER}}=\mathrm{f}\left({\mathrm{G}}_{\mathrm{SURROUND}}\right)\\ {\hat{\mathrm{H}}}_{\mathrm{CENTER}}={\mathrm{H}}_{\mathrm{CENTER}}\\ {\hat{\mathrm{J}}}_{\mathrm{CENTER}}={\mathrm{G}}_{\mathrm{CENTER}}+{\mathrm{v}}_{\mathrm{J}\mathrm{G}}\end{array}$$
  
  where f(·) is an average operator, whose inputs are the surrounding pixels (H_SURROUND or G_SURROUND), along the direction provided by the zipper detector block; $${\mathrm{v}}_{\mathrm{JG}}=\mathrm{median\hspace{1em}}\{{\mathrm{J}}_{\mathrm{ij}}-{\mathrm{G}}_{\mathrm{ij}}|(\mathrm{i},\mathrm{j})\in \mathrm{\Re }\}$$
  
  G is green (G), H is red (R) or blue (B) and J is blue (B) or red (R), respectively, and
  
  is the support of the median filter which covers 3x3 pixels. Other support dimensions, such as e.g. 5x5 pixels can be used
• As in the false colors removal algorithm, the new values Ĝ_CENTER, Ĥ_CENTER and Ĵ_CENTER will replace the colors of the central pixel.
• Figure 3 is an exemplary block diagram of circuitry adapted to implement the rules presented above (i.e. eqs. 1 to 4 above).
• Specifically, in Figure 3, the selector block 60, according to the central pixel channel, activates one of the three possible flow paths (H, G, J), enabling the blocks involved in the related calculation, while two blocks designated 60a and 60b compute the two median values H - G (v_HG), and J - G (v_JG), respectively.
• The blocks designated 70 calculate two directional average values as average of some surrounding pixels of the central pixel. The blocks designated 70a, 70b, 70c, and 70d calculate the values H - v_HG, G + v_HG, G + v_JG, and J - v_JG, respectively.
• Multiplexers designated 80 each select (under control signals provided via lines 90 by the zipper detector block 30) a respective output out of a set of three possibilities inputs, while the blocks designated 80a and 80b produce the values G' + v_JG and G' + v_HG. The various blocks just described thus produce the signals H, G', J'; H', G, J'; and H', G', J over output lines designated by corresponding references in Figure 3.
• Without prejudice to the underlying principles of the invention, the details and the embodiments may vary, also appreciably, with reference to what has been described by way of example only, without departing from the scope of the invention as defined by the annexed claims.

Claims (37)

1. A method of processing color image signals arranged in pixels, each pixel having associated detected (10) color information for a first color as well as undetected filled-in (20) color information for at least a second and a third color, whereby said images are exposed to the presence of false color and zipper effect artifacts, the method including the steps of:

   - checking (30) said images for the presence of zipper effect artifacts, and

   - i) if said checking (30) reveals the presence of zipper effect artifacts, applying a zipper effect removal process (50) to said image signals;

   - ii) if said checking (30) fails to reveal the presence of zipper effect artifacts, applying a false color removal process (40) to said image signals,

   whereby said false color and zipper effect artifacts are both reduced by adaptively using said zipper effect removal process (50) and said false color removal process (40).
2. The method of claim 1, characterized in that said checking (30) for the presence of zipper effect artifacts includes checking the conditions that inter-channel differences between one of said one second and third colors and said first color are almost constant along one direction of said images and not constant in the direction orthogonal thereto.
3. The method of claim 2, characterized in that said checking (30) said conditions includes the step of evaluating the similarity of said inter-channel differences against a given threshold, whereby increasing said threshold increased the amount of zipper effect removed by said zipper effect removal process (50).
4. The method of claim 2, characterized in that said checking (30) for the presence of zipper effect artifacts includes checking, along said direction having not constant inter-channel differences, whether a given pixel:

   - belongs to a line separating a dark region and a light one in the respective image;

   - belongs to a dark line within a light background in the respective image;

   - belongs to a light line within a dark background in the respective image.
5. The method of any of claims 2 to 4, characterized in that said checking (30) for the presence of zipper effect artifacts includes checking that:

   - along the a first direction passing through a given central pixel, the inter-channel difference between one of said second and third colors and said first color is substantially constant;

   - along a second direction, orthogonal to said first direction, the inter-channel difference between one of said second and third colors and said first color is not constant;

   - the trend of the central pixel channel in said second direction is increasing or decreasing, or there is a minimum or a maximum in said given central pixel.
6. The method of any of the previous claims,
   characterized in that said step of checking for the presence of zipper effect artifacts is performed on a mask of pixels.
7. The method of claim 6, characterized in that said step of checking for the presence of zipper effect artifacts is performed on a mask of pixels selected out of a 5x5 mask of pixels and a 3x3 mask of pixels.
8. The method of claim 1, characterized in that both said zipper effect removal process (50) and said false color removal process (40) include the steps of:

   - for each pixel, leaving unchanged said detected (10) color information for a first color;

   - for each pixel, selecting a given neighborhood of said each pixel,

   - producing at least one filtered interchannel difference (v_JG; v_HG) over said given neighborhood of said each pixel, said filtered interchannel difference involving one of said undetected filled-in (20) color information for at least a second and a third color,

   - using said at least one filtered interchannel difference (v_JG; v_HG) for performing both said zipper effect removal process (50) and said false color removal process (40).
9. A method of removing zipper effect artifacts from color image signals arranged in pixels, each pixel having associated detected (10) color information for a first color as well as undetected filled-in (20) color information for at least a second and a third color, the method including the steps of:

   - for each pixel, leaving unchanged said detected (10) color information for a first color;

   - for each pixel, selecting a given neighborhood of said each pixel,

   - producing at least one filtered interchannel difference (v_JG) over said given neighborhood of said each pixel, said filtered interchannel difference involving one of said at least a second and a third color,

   - adding to the color corresponding to said one of said undetected filled-in (20) color information, said filtered inter-channel difference (v_JG),

   - calculating the other of said second and third colors as the average of some surrounding pixels of said each pixel.
10. The method of claim 9, characterized in that it includes the step of selecting said surrounding pixels along a direction wherein said at least one inter-channel difference is substantially constant.
11. The method of either of claims 9 or 10,
    characterized in that said least one filtered interchannel difference (v_HG; v_JG) over said given neighborhood of said each pixel is a median filtered interchannel difference.
12. The method of any of claims 9 to 11,
    characterized in that said color image signals are red-green-blue (RGB) image signals, and said zipper effect removal process (50) includes the steps of:

    - if said first color is green (G), setting: $${\widehat{\mathrm{G}}}_{\mathrm{CENTER}}={\mathrm{G}}_{\mathrm{CENTER}}$$

    

    $${\widehat{\mathrm{H}}}_{\mathrm{CENTER}}=\mathrm{f}\left({\mathrm{H}}_{\mathrm{SURROUND}}\right)$$

    

    $${\widehat{\mathrm{J}}}_{\mathrm{CENTER}}={\mathrm{G}}_{\mathrm{CENTER}}+{\mathrm{v}}_{\mathrm{J}\mathrm{G}}$$

    

    - if said first color is red (R) or blue (B), setting: $${\widehat{\mathrm{G}}}_{\mathrm{CENTER}}=\mathrm{f}\left({\mathrm{G}}_{\mathrm{SURROUND}}\right)$$

    

    $${\widehat{\mathrm{H}}}_{\mathrm{CENTER}}={\mathrm{H}}_{\mathrm{CENTER}}$$

    

    $${\widehat{\mathrm{J}}}_{\mathrm{CENTER}}={\widehat{\mathrm{G}}}_{\mathrm{CENTER}}+{\mathrm{v}}_{\mathrm{J}\mathrm{G}}$$

    

    where:

    G is green, H is red (R) or blue (B) and J is blue (B) or red (R), respectively;

    _···CENTER designates said each pixel;

    ^ denotes the signals resulting from the zipper effect removal process (50) and replacing the colors at said each pixel;

    f(·) is an average operator, whose inputs are the pixels (H_SURROUND or G_SURROUND) surrounding said each pixel, along the direction provided by zipper detection;

    v_JG= median {J_ij-G_ij|(i,j)∈

    

    } is said at least one interchannel difference;

    and

    

    is the support of a median filter which covers said given neighborhood.
13. The method of any of claims 9 to 12, characterized in that said given neighborhood is selected out of a neighborhood including 3x3 pixels and a neighborhood including 5x5 pixels.
14. A method of removing false color artifacts from color image signals arranged in pixels, each pixel having associated detected (10) color information for a first color as well as undetected filled-in (20) color information for at least a second and a third color, the method including the steps of:

    - for each pixel, leaving unchanged said detected (10) color information for a first color;

    - selecting a given neighborhood of said each pixel,

    - producing at least two filtered interchannel differences (v_HG; v_JG) over said given neighborhood of said each pixel, said interchannel differences involving said at least a second and a third color, and

    - effecting said false color removal (40) using both said interchannel differences, whereby the spectral correlation between color planes is taken into account in said false color removal process (40).
15. The method of Claim 14, characterized in that both said two filtered interchannel differences (v_HG; v_JG) are used for false color removal (40) from both said undetected filled-in (20) color information for at least a second and a third color.
16. The method of either of Claims 14 or 15,
    characterized in that it includes the steps of producing said filtered interchannel differences (v_HG; v_JG) as median filtered interchannel differences (v_HG; v_JG) over said given neighborhood of said each pixel.
17. The method of any of Claims 14 to 16,
    characterized in that said color image signals are red-green-blue (RGB) image signals and said false color removal process (40) includes the steps of:

    - if said first color is green (G), setting: $${\widehat{\mathrm{H}}}_{\mathrm{CENTER}}={\mathrm{G}}_{\mathrm{CENTER}}+{\mathrm{v}}_{\mathrm{H}\mathrm{G}}$$

    

    $${\widehat{\mathrm{G}}}_{\mathrm{CENTER}}={\mathrm{G}}_{\mathrm{CENTER}}$$

    

    $${\widehat{\mathrm{J}}}_{\mathrm{CENTER}}={\mathrm{G}}_{\mathrm{CENTER}}+{\mathrm{v}}_{\mathrm{J}\mathrm{G}}$$

    

    - if said first color is either red (R) or blue (B), setting: $${\widehat{\mathrm{H}}}_{\mathrm{CENTER}}={\mathrm{H}}_{\mathrm{CENTER}}$$

    

    $${\widehat{\mathrm{G}}}_{\mathrm{CENTER}}={\mathrm{H}}_{\mathrm{CENTER}}-{\mathrm{v}}_{\mathrm{H}\mathrm{G}}$$

    

    $${\widehat{\mathrm{J}}}_{\mathrm{CENTER}}={\mathrm{H}}_{\mathrm{CENTER}}-{\mathrm{v}}_{\mathrm{H}\mathrm{G}}+{\mathrm{v}}_{\mathrm{J}\mathrm{G}}$$

    

    where:

    G is green (G), H is red (R) or blue (B) and J is blue (B) or red (R), respectively;

    _{··· CENTER} designates said each pixel;

    ^ denotes the signals resulting from the false color removal process (40) and replacing the colors at said each pixel; $${\mathrm{v}}_{\mathrm{H}\mathrm{G}}=\mathrm{median}\left\{{\mathrm{H}}_{\mathrm{i}\mathrm{j}}-{\mathrm{G}}_{\mathrm{i}\mathrm{j}}|\left(\mathrm{i},\mathrm{j}\right)\in \mathrm{\Re }\right\},$$

    

    $${\mathrm{v}}_{\mathrm{J}\mathrm{G}}=\mathrm{median}\left\{{\mathrm{J}}_{\mathrm{i}\mathrm{j}}-{\mathrm{G}}_{\mathrm{i}\mathrm{j}}|\left(\mathrm{i},\mathrm{j}\right)\in \mathrm{\Re }\right\}$$

    

    are said interchannel differences;
    , and
    

    

    is the support of a median filter which covers a local window surrounding said each pixel.
18. The method of Claim 17, characterized in that said local window is selected out of windows including 3x3 pixels and 5X5 pixels.
19. A system for processing color image signals arranged in pixels, each pixel having associated detected (10) color information for a first color as well as undetected filled-in (20) color information for at least a second and a third color, whereby said images are exposed to the presence of false color and zipper effect artifacts, the system including:

    - a zipper detector (30) for checking said images for the presence of zipper effect artifacts,

    - zipper effect removal module (50) configured for applying a zipper effect removal process (50) to said image signals if said checking by said zipper detector (30) reveals the presence of zipper effect artifacts; and

    - a false color removal module (40) configured for applying a false color removal process (40) to said image signals if said checking by said zipper detector (30) fails to reveal the presence of zipper effect artifacts;

    the arrangement being such that said false color and zipper effect artifacts are both reduced by adaptively using said zipper effect removal process (50) and said false color removal process (40).
20. The system of claim 19, characterized in that said zipper detector (30) is configured for checking the conditions that inter-channel differences between one of said one second and third colors and said first color are almost constant along one direction of said images and not constant in the direction orthogonal thereto.
21. The system of claim 20, characterized in that said zipper detector (30) is configured for evaluating the similarity of said inter-channel differences against a given threshold, whereby increasing said threshold increased the amount of zipper effect removed by said zipper effect removal process (50).
22. The system of claim 20, characterized in that said zipper detector (30) is configured for checking, along said direction having not constant inter-channel differences, whether a given pixel:

    - belongs to a line separating a dark region and a light one in the respective image;

    - belongs to a dark line within a light background in the respective image;

    - belongs to a light line within a dark background in the respective image.
23. The system of any of claims 20 to 22,
    characterized in that said zipper detector (30) is configured for checking that:

    - along the a first direction passing through a given central pixel, the inter-channel difference between one of said second and third colors and said first color is substantially constant;

    - along a second direction, orthogonal to said first direction, the inter-channel difference between one of said second and third colors and said first color is not constant;

    - the trend of the central pixel channel in said second direction is increasing or decreasing, or there is a minimum or a maximum in said given central pixel.
24. The system of any of the previous claims 19 to 25, characterized in that said zipper detector (30) is configured for checking for the presence of zipper effect artifacts on a mask of pixels.
25. The system of claim 24, characterized in that said zipper detector (30) is configured for checking for the presence of zipper effect artifacts on a mask of pixels selected out of a 5x5 mask of pixels and a 3x3 mask of pixels.
26. The system of claim 19, characterized in that said zipper effect removal module (50) and said false color removal module (40) are jointly configured for:

    - for each pixel, leaving unchanged said detected (10) color information for a first color;

    - for each pixel, selecting a given neighborhood of said each pixel,

    - producing at least one filtered interchannel difference (v_JG; v_HG) over said given neighborhood of said each pixel, said filtered interchannel difference involving one of said undetected filled-in (20) color information for at least a second and a third color,

    - using said at least one filtered interchannel difference (v_JG; v_HG) for performing both said zipper effect removal process (50) and said false color removal process (40).
27. A system of removing zipper effect artifacts from color image signals arranged in pixels, each pixel having associated detected (10) color information for a first color as well as undetected filled-in (20) color information for at least a second and a third color, the system including a zipper effect removal module (50) configured for:

    - for each pixel, leaving unchanged said detected (10) color information for a first color;

    - for each pixel, selecting a given neighborhood of said each pixel,

    - producing at least one filtered interchannel difference (v_JG) over said given neighborhood of said each pixel, said filtered interchannel difference involving one of said at least a second and a third color,

    - adding to the color corresponding to said one of said undetected filled-in (20) color information, said filtered inter-channel difference (v_JG),

    - calculating the other of said second and third colors as the average of some surrounding pixels of said each pixel.
28. The system of claim 27, characterized in that said zipper effect removal module (50) is configured for selecting said surrounding pixels along a direction
    wherein said at least one inter-channel difference is substantially constant.
29. The system of either of claims 27 or 28,
    characterized in that said least one filtered interchannel difference (v_HG; v_JG) over said given neighborhood of said each pixel is a median filtered interchannel difference.
30. The system of any of claims 27 to 29,
    characterized in that said color image signals are red-green-blue (RGB) image signals, and said zipper effect removal module (50) is configured for:

    - if said first color is green (G), setting: $${\widehat{\mathrm{G}}}_{\mathrm{CENTER}}={\mathrm{G}}_{\mathrm{CENTER}}$$

    

    $${\widehat{\mathrm{H}}}_{\mathrm{CENTER}}=\mathrm{f}\left({\mathrm{H}}_{\mathrm{SURROUND}}\right)$$

    

    $${\widehat{\mathrm{J}}}_{\mathrm{CENTER}}={\mathrm{G}}_{\mathrm{CENTER}}+{\mathrm{v}}_{\mathrm{J}\mathrm{G}}$$

    

    - if said first color is red (R) or blue (B), setting: $${\widehat{\mathrm{G}}}_{\mathrm{CENTER}}=\mathrm{f}\left({\mathrm{G}}_{\mathrm{SURROUND}}\right)$$

    

    $${\widehat{\mathrm{H}}}_{\mathrm{CENTER}}={\mathrm{H}}_{\mathrm{CENTER}}$$

    

    $${\widehat{\mathrm{J}}}_{\mathrm{CENTER}}={\widehat{\mathrm{G}}}_{\mathrm{CENTER}}+{\mathrm{v}}_{\mathrm{J}\mathrm{G}}$$

    

    where:

    G is green, H is red (R) or blue (B) and J is blue (B) or red (R), respectively;

    _···CENTER designates said each pixel;

    ^ denotes the signals resulting from the zipper effect removal process (50) and replacing the colors at said each pixel;

    f(·) is an average operator, whose inputs are the pixels (H_SURROUND or G_SURROUND) surrounding said each pixel, along the direction provided by zipper detection;

    v_JG = median {J_ij-G_ij|(i,j)∈

    

    } is said at least one interchannel difference;

    and

    

    is the support of a median filter which covers said given neighborhood.
31. The system of any of claims 27 to 30,
    characterized in that zipper effect removal module (50) is configured for selecting said given neighborhood out of a neighborhood including 3x3 pixels and a neighborhood including 5x5 pixels.
32. A system of removing false color artifacts from color image signals arranged in pixels, each pixel having associated detected (10) color information for a first color as well as undetected filled-in (20) color information for at least a second and a third color, the system including a false color removal module (40) configured for:

    - for each pixel, leaving unchanged said detected (10) color information for a first color;

    - selecting a given neighborhood of said each pixel,

    - producing at least two filtered interchannel differences (v_HG; v_JG) over said given neighborhood of said each pixel, said interchannel differences involving said at least a second and a third color, and

    - effecting said false color removal (40) using both said interchannel differences, whereby the spectral correlation between color planes is taken into account in said false color removal process (40).
33. The system of Claim 32, characterized in that said false color removal module (40) is configured for using both said two filtered interchannel differences (v_HG; v_JG) for false color removal (40) from both said undetected filled-in (20) color information for at least a second and a third color.
34. The system of either of Claims 32 or 33,
    characterized in that said false color removal module (40) is configured for producing said filtered interchannel differences (v_HG; v_JG) as median filtered interchannel differences (v_HG; v_JG) over said given neighborhood of said each pixel.
35. The system of any of Claims 32 to 34,
    characterized in that said color image signals are red-green-blue (RGB) image signals and said false color removal module (40) is configured for:

    - if said first color is green (G), setting: $${\widehat{\mathrm{H}}}_{\mathrm{CENTER}}={\mathrm{G}}_{\mathrm{CENTER}}+{\mathrm{v}}_{\mathrm{H}\mathrm{G}}$$

    

    $${\widehat{\mathrm{G}}}_{\mathrm{CENTER}}={\mathrm{G}}_{\mathrm{CENTER}}$$

    

    $${\widehat{\mathrm{J}}}_{\mathrm{CENTER}}={\mathrm{G}}_{\mathrm{CENTER}}+{\mathrm{v}}_{\mathrm{J}\mathrm{G}}$$

    

    - if said first color is either red (R) or blue (B), setting: $${\widehat{\mathrm{H}}}_{\mathrm{CENTER}}={\mathrm{H}}_{\mathrm{CENTER}}$$

    

    $${\widehat{\mathrm{G}}}_{\mathrm{CENTER}}={\mathrm{H}}_{\mathrm{CENTER}}-{\mathrm{v}}_{\mathrm{H}\mathrm{G}}$$

    

    $${\widehat{\mathrm{J}}}_{\mathrm{CENTER}}={\mathrm{H}}_{\mathrm{CENTER}}-{\mathrm{v}}_{\mathrm{H}\mathrm{G}}+{\mathrm{v}}_{\mathrm{J}\mathrm{G}}$$

    

    where:

    G is green (G), H is red (R) or blue (B) and J is blue (B) or red (R), respectively;

    _{··· CENTER} designates said each pixel;

    ^ denotes the signals resulting from the false color removal process (40) and replacing the colors at said each pixel; $${\mathrm{v}}_{\mathrm{H}\mathrm{G}}=\mathrm{median}\left\{{\mathrm{H}}_{\mathrm{i}\mathrm{j}}-{\mathrm{G}}_{\mathrm{i}\mathrm{j}}|\left(\mathrm{i},\mathrm{j}\right)\in \mathrm{\Re }\right\},$$

    

    $${\mathrm{v}}_{\mathrm{J}\mathrm{G}}=\mathrm{median}\left\{{\mathrm{J}}_{\mathrm{i}\mathrm{j}}-{\mathrm{G}}_{\mathrm{i}\mathrm{j}}|\left(\mathrm{i},\mathrm{j}\right)\in \mathrm{\Re }\right\}$$

    

    are said interchannel differences;
    , and
    

    

    is the support of a median filter which covers a local window surrounding said each pixel.
36. The system of Claim 35, characterized in that said local window is selected out of windows including 3x3 pixels and 5X5 pixels.
37. A computer program product, loadable in the memory of at least one computer and including software code portions for performing the steps of the method of any of Claims 1 to 18.

Images